<feed xmlns='http://www.w3.org/2005/Atom'>
<title>Tianocore/edk2.git, branch dependabot/github_actions/github/codeql-action-4.39.3</title>
<subtitle>EDK II (mirror)</subtitle>
<id>https://git.radix-linux.su/Tianocore/edk2.git/atom?h=dependabot%2Fgithub_actions%2Fgithub%2Fcodeql-action-4.39.3</id>
<link rel='self' href='https://git.radix-linux.su/Tianocore/edk2.git/atom?h=dependabot%2Fgithub_actions%2Fgithub%2Fcodeql-action-4.39.3'/>
<link rel='alternate' type='text/html' href='https://git.radix-linux.su/Tianocore/edk2.git/'/>
<updated>2026-11-09T13:15:53+00:00</updated>
<entry>
<title>GitHub Action: Bump github/codeql-action from 4.37.4 to 4.39.3</title>
<updated>2026-11-09T13:15:53+00:00</updated>
<author>
<name>dependabot[bot]</name>
<email>49699333+dependabot[bot]@users.noreply.github.com</email>
</author>
<published>2026-11-09T13:15:53+00:00</published>
<link rel='alternate' type='text/html' href='https://git.radix-linux.su/Tianocore/edk2.git/commit/?id=854563532e0ba64fb75f36cd6379625bcdaf718c'/>
<id>urn:sha1:854563532e0ba64fb75f36cd6379625bcdaf718c</id>
<content type='text'>
Bumps [github/codeql-action](https://github.com/github/codeql-action) from 4.37.4 to 4.39.3.
- [Release notes](https://github.com/github/codeql-action/releases)
- [Changelog](https://github.com/github/codeql-action/blob/main/CHANGELOG.md)
- [Commits](https://github.com/github/codeql-action/compare/v4.37.4...v4.39.3)

---
updated-dependencies:
- dependency-name: github/codeql-action
  dependency-version: 4.39.3
  dependency-type: direct:production
  update-type: version-update:semver-minor
...

Signed-off-by: dependabot[bot] &lt;support@github.com&gt;</content>
</entry>
<entry>
<title>MdeModulePkg: Use unsigned memory type bounds checks</title>
<updated>2026-11-06T01:42:08+00:00</updated>
<author>
<name>Mattia Giambirtone</name>
<email>nocturn9x@nocturn9x.space</email>
</author>
<published>2026-11-05T12:16:01+00:00</published>
<link rel='alternate' type='text/html' href='https://git.radix-linux.su/Tianocore/edk2.git/commit/?id=b9d8871d3432048a47371e39f5b7413a40e4a774'/>
<id>urn:sha1:b9d8871d3432048a47371e39f5b7413a40e4a774</id>
<content type='text'>
Some compilers treat EFI_MEMORY_TYPE as signed. OS-reserved memory
types with bit 31 set can therefore pass a signed comparison against
EfiMaxMemoryType and index outside the memory type statistics array.

Cast the memory type to UINT32 before checking the array bound in
CoreTerminateMemoryMap and InternalBuildMemoryAllocationHob, matching
the existing guards in the page, pool, and memory bin code.

Fixes: https://github.com/tianocore/edk2/issues/13451
Signed-off-by: Mattia Giambirtone &lt;nocturn9x@nocturn9x.space&gt;
</content>
</entry>
<entry>
<title>OvmfPkg/PlatformCI: Verify serial console and PS/2 keyboard in boot test</title>
<updated>2026-11-05T09:37:23+00:00</updated>
<author>
<name>Kevin Loughlin</name>
<email>kevinloughlin@google.com</email>
</author>
<published>2026-11-04T20:28:48+00:00</published>
<link rel='alternate' type='text/html' href='https://git.radix-linux.su/Tianocore/edk2.git/commit/?id=e788c5a39dff773a545a728a01f07ae6bf614729'/>
<id>urn:sha1:e788c5a39dff773a545a728a01f07ae6bf614729</id>
<content type='text'>
Currently, FlashRomImage() in PlatformBuildLib.py boots OVMF in QEMU
with -debugcon stdio and a startup.nsh script, but only checks the QEMU
process exit code. The UEFI Shell still executes startup.nsh and shuts
down cleanly even when ConIn and ConOut fail to connect on the ISA/LPC
bridge. Because -serial output is neither captured nor inspected,
regressions in ISA serial (PNP0501) or PS/2 keyboard (PNP0303) console
initialization go undetected in CI.

When MAKE_STARTUP_NSH=TRUE:
1. Redirect QEMU serial output to &lt;BUILD_OUTPUT_BASE&gt;/serial.log via
   -serial file:&lt;path&gt;.
2. Run "dh -d -v -p SimpleTextIn" in startup.nsh before "reset -s".
3. After QEMU exits, verify that serial.log contains "Serial(0x0)"
   (the UefiDevicePathLib text representation of EISA_PNP_ID(0x0501)),
   "Acpi(PNP0303,0x0)", and "BOOT SUCCESS !!!".

Signed-off-by: Kevin Loughlin &lt;kevinloughlin@google.com&gt;
</content>
</entry>
<entry>
<title>OvmfPkg/PlatformBootManagerLib: Enable PCI decoding on LPC bridge</title>
<updated>2026-11-05T09:37:23+00:00</updated>
<author>
<name>Kevin Loughlin</name>
<email>kevinloughlin@google.com</email>
</author>
<published>2026-11-04T16:54:01+00:00</published>
<link rel='alternate' type='text/html' href='https://git.radix-linux.su/Tianocore/edk2.git/commit/?id=e4771cff158b8bcbe3ea9f79e1809f0a48877bfc'/>
<id>urn:sha1:e4771cff158b8bcbe3ea9f79e1809f0a48877bfc</id>
<content type='text'>
Commit 01bd046e7b4b ("OvmfPkg/PlatformBootManagerLib: Remove blanket PCI
device enablement") removed the unconditional PciIo-&gt;Attributes()
EFI_PCI_DEVICE_ENABLE call across all PCI devices in
DetectAndPreparePlatformPciDevicePath().

Unlike PCI-to-PCI bridges (which PciBusDxe enables automatically in
StartPciDevicesOnBridge()), the PCI-to-ISA (LPC) bridge requires
EFI_PCI_DEVICE_ENABLE in DetectAndPreparePlatformPciDevicePath() for two
reasons:
1. SioBusDriverBindingSupported() in OvmfPkg/SioBusDxe is the only
   DriverBindingSupported() implementation in edk2 that checks
   Pci.Hdr.Command. It requires PCI I/O and memory space decoding
   ((Pci.Hdr.Command &amp; 0x03) == 0x03) to already be enabled on the
   bridge before binding, so ISA serial (PNP0501) and PS/2 keyboard
   (PNP0303) console initialization fail without it.
2. Guest operating systems access legacy I/O ports behind the LPC/ISA
   bridge directly without a dedicated PCI driver for the LPC bridge
   device itself.

Re-enable EFI_PCI_DEVICE_ENABLE specifically when the LPC bridge is
detected in DetectAndPreparePlatformPciDevicePath() in
PlatformBootManagerLib and PlatformBootManagerLibBhyve.

Suggested-by: Ard Biesheuvel &lt;ardb+tianocore@kernel.org&gt;
Signed-off-by: Kevin Loughlin &lt;kevinloughlin@google.com&gt;
</content>
</entry>
<entry>
<title>ArmPkg/ArmPsciMpServicesDxe: Clean the image to the PoC after init</title>
<updated>2026-11-05T07:13:36+00:00</updated>
<author>
<name>Ard Biesheuvel</name>
<email>ardb@kernel.org</email>
</author>
<published>2026-11-01T15:11:37+00:00</published>
<link rel='alternate' type='text/html' href='https://git.radix-linux.su/Tianocore/edk2.git/commit/?id=561c006ac6bb962d8e53b369b6a6a01d2d9ae579'/>
<id>urn:sha1:561c006ac6bb962d8e53b369b6a6a01d2d9ae579</id>
<content type='text'>
The APs load gTcr, gMair, gTtbr0 and gHcr in ApEntryPoint before they
enable their MMU and caches, i.e., using non-cacheable accesses that are
not coherent with the BSP's data cache. The driver cleans its image to
the PoC in its entry point to cater for this, but it does so before
calling MpServicesInitialize(), which is where these variables are
assigned. Of these, only gHcr is cleaned explicitly afterwards, and so
gTcr, gMair and gTtbr0 only make it to the PoC if they happen to share a
cache line with gHcr. This is the case in an ArmVirtQemu DEBUG build
with GCC 15, for instance, but it is not guaranteed in general.

Note that this cannot be reproduced under TCG, which does not model
caches, nor under KVM on hosts with FEAT_S2FWB, where the MMU-off
accesses are cacheable.

So move the cache maintenance on the entire image to after the call to
MpServicesInitialize(), which covers all the global variables accessed
by the APs, and drop the explicit cleaning of individual variables, as
it is now redundant.

While at it, drop the cleaning of the gProcessorIDs array as well: it is
allocated from pool memory, and so it is not covered by the cache
maintenance on the image, but the APs only access it after enabling
their MMU and caches, i.e., using cacheable accesses that are coherent
with the BSP's data cache.

Signed-off-by: Ard Biesheuvel &lt;ardb@kernel.org&gt;
</content>
</entry>
<entry>
<title>Revert "MdeModulePkg/PciBusDxe: Preserve physical BAR numbering"</title>
<updated>2026-11-05T04:06:27+00:00</updated>
<author>
<name>Aaron Pop</name>
<email>aaronpop@microsoft.com</email>
</author>
<published>2026-10-23T22:34:42+00:00</published>
<link rel='alternate' type='text/html' href='https://git.radix-linux.su/Tianocore/edk2.git/commit/?id=087a07536de8b8561c7ae5ed974e60c80a9ed4ae'/>
<id>urn:sha1:087a07536de8b8561c7ae5ed974e60c80a9ed4ae</id>
<content type='text'>
This reverts commit 0e7dad44bae663193b580adeb9d942c2559166f6.

Signed-off-by: Aaron Pop &lt;aaronpop@microsoft.com&gt;
</content>
</entry>
<entry>
<title>Revert "MdeModulePkg/PciBusDxe: Add Google Tests for BAR parsing"</title>
<updated>2026-11-05T04:06:27+00:00</updated>
<author>
<name>Aaron Pop</name>
<email>aaronpop@microsoft.com</email>
</author>
<published>2026-10-23T22:31:12+00:00</published>
<link rel='alternate' type='text/html' href='https://git.radix-linux.su/Tianocore/edk2.git/commit/?id=f4e2af190443b116a327ba017f8b2612b01fbb08'/>
<id>urn:sha1:f4e2af190443b116a327ba017f8b2612b01fbb08</id>
<content type='text'>
This reverts commit 0d61209231b2ec50ca6f6b96825baaa07c280854.

Signed-off-by: Aaron Pop &lt;aaronpop@microsoft.com&gt;
</content>
</entry>
<entry>
<title>MdeModulePkg/CapsulePei: Print 64-bit capsule addresses in DEBUG logs</title>
<updated>2026-11-05T03:28:06+00:00</updated>
<author>
<name>Phil Noh</name>
<email>Phil.Noh@amd.com</email>
</author>
<published>2026-10-30T20:32:49+00:00</published>
<link rel='alternate' type='text/html' href='https://git.radix-linux.su/Tianocore/edk2.git/commit/?id=965f4c22530d43147a456977705545874674aa0f'/>
<id>urn:sha1:965f4c22530d43147a456977705545874674aa0f</id>
<content type='text'>
Three DEBUG messages in CapsulePei print capsule addresses: the coalesce
memory range, the free memory range, and the capsule copy source and
destination. All three used the "0x%8X" format specifier with UINTN
arguments. On X64, this truncates the printed value to 32 bits, so when
the capsule coalesce buffer is allocated above 4GB, the logs show a
misleading address.

Change these messages to use the "0x%lX" specifier and cast the
arguments to UINT64 so that the full 64-bit address and size are printed
on X64 while remaining correct on IA32.

Signed-off-by: Phil Noh &lt;Phil.Noh@amd.com&gt;
</content>
</entry>
<entry>
<title>NetworkPkg/IScsiDxe: Check bounds before root path access</title>
<updated>2026-11-04T19:35:46+00:00</updated>
<author>
<name>Aaron Pop</name>
<email>aaronpop@microsoft.com</email>
</author>
<published>2026-09-09T20:37:54+00:00</published>
<link rel='alternate' type='text/html' href='https://git.radix-linux.su/Tianocore/edk2.git/commit/?id=6e3d7df43d56d3870b06828d94348371f752e282'/>
<id>urn:sha1:6e3d7df43d56d3870b06828d94348371f752e282</id>
<content type='text'>
CodeQL static analysis reports out-of-bounds reads in the iSCSI DHCP
root path parsing loops. The loop conditions read TmpStr[Index] before
checking Index against Length.

When a delimiter is absent before the end of the root path, Index
reaches Length and the old conditions perform a one-past-end read before
terminating. DHCP root paths are externally supplied, so the parser must
validate the index before accessing the buffer.

Reorder the conditions in the IPv4 and IPv6 parsing paths so the bounds
check short-circuits the buffer access. This prevents the out-of-bounds
reads and resolves the CodeQL static scanning failures.

Signed-off-by: Aaron Pop &lt;aaronpop@microsoft.com&gt;
</content>
</entry>
<entry>
<title>NetworkPkg/IScsiDxe: Align attempt count types</title>
<updated>2026-11-04T19:35:46+00:00</updated>
<author>
<name>Aaron Pop</name>
<email>aaronpop@microsoft.com</email>
</author>
<published>2026-09-02T21:26:43+00:00</published>
<link rel='alternate' type='text/html' href='https://git.radix-linux.su/Tianocore/edk2.git/commit/?id=2bcdc4c1ab5292b153b592eb11711f4025688545'/>
<id>urn:sha1:2bcdc4c1ab5292b153b592eb11711f4025688545</id>
<content type='text'>
Define the attempt count and loop index as UINT8 to match
PcdMaxIScsiAttemptNumber and the stored attempt index type. This avoids
comparing the attempt index against a wider integer type.

Signed-off-by: Aaron Pop &lt;aaronpop@microsoft.com&gt;
</content>
</entry>
</feed>
